10  DSC: Lymphoma vs GBM

I would like to differentiate between “Glioblastoma” vs “Primary CNS lymphoma” using advanced brain MRI technique such as DSC (rCBV), DCE (k-trans), and MR spectroscopy.

In one line: Combine perfusion and metabolic fingerprints—low rCBV + very high Ktrans/extraction-fraction + prominent lipid/lactate peaks in a non-necrotic mass → PCNSL, whereas high heterogenous rCBV ± ring necrosis + only moderate permeability + variable lipids confined to necrotic core → GBM.


10.1 1 | Dynamic Susceptibility-Contrast (DSC) Perfusion

10.1.1 How to acquire

  • Pre-load + leakage-corrected processing to neutralise T1-shine-through, then normalise the max-rCBV to contralateral white matter. (PMC)
  • Place a small ROI on the highest-CBV “hot spot” (avoid visible vessels/necrosis) and a larger ROI for mean-rCBV; record the percentage-signal-recovery (PSR) curve as well. (EJ Radiology)

10.1.2 Typical findings

Metric Glioblastoma (GBM) Primary CNS Lymphoma (PCNSL) Practical cut-off
Max- or mean-rCBV 3–6 (mean ≈ 4.9 ± 2.2) (PubMed) 0.8–2.0 (mean ≈ 1.7 ± 0.6) (PubMed, PMC) < 1.5–1.7 ⇒ strongly favours PCNSL (PMC)
PSR (signal recovery) Low/variable (due to neo-vascular leak) (PMC) High PSR (> 90 %)—reflects slow wash-out and intact lumina (PubMed)
Extraction fraction (from DSC leakage models) Lower Higher—even in hyper-vascular outliers (ajnr.org) > 0.25 suggests PCNSL

Why? GBM shows florid angiogenesis with many patent vessels, hence high blood volume; PCNSL grows angiocentrically, destroys endothelium, but recruits few new vessels → low CBV yet marked contrast leak. (PMC)


10.2 2 | Dynamic Contrast-Enhanced (DCE) Perfusion

10.2.1 Acquisition pearls

  • 3-D spoiled gradient-echo every 2–4 s for ≥5 min; Tofts model for Ktrans, ve, Kep.
  • Use the same ROI strategy as DSC and map peritumoral rim separately. (PubMed)

10.2.2 Pharmacokinetic signatures

Parameter GBM PCNSL
Ktrans (min⁻¹) 0.2–0.6 (moderate) (PubMed) 0.3–0.8 (often higher) (PubMed, PubMed)
Kep (min⁻¹) High in peritumoural zone (disrupted BBB) (PubMed) Lower
ve Larger heterogeneous extracellular space Smaller, homogeneous
Diagnostic ROC Kep peritumoural AUC ≈ 0.90; Ktrans core AUC ≈ 0.82 (PubMed, ajnr.org)

Interpretation tip: a mass with low rCBV but high Ktrans is classic for lymphoma; high rCBV with only moderate Ktrans suggests GBM.


10.3 3 | Proton MR Spectroscopy (¹H-MRS)

Spectral marker GBM PCNSL
Cho/Cr ratio 2–4; elevated in tumour & necrotic rim > 2.6 (mean ≈ 3–4) (PMC)
Lipids/Lactate Prominent only in necrotic core Massive peaks even in solid, non-necrotic tissue (PubMed, The Journal of Neurosurgery)
Glu/Cr Lower Higher (cut-off ≈ 2.5) (PubMed)
NAA Markedly decreased Markedly decreased
Myo-inositol Variable Often low

Pearls

  • A large lipid peak without frank necrosis is almost pathognomonic for PCNSL. (The Journal of Neurosurgery)
  • Elevated glycine & alanine favour GBM but are not specific.

10.4 4 | Step-by-Step Workflow for Unknown Enhancing Mass

  1. Conventional MRI: if homogeneous enhancement + diffusion restriction → think PCNSL; if ring-necrosis → think GBM.

  2. DSC

    • rCBV < 1.5 → go to DCE/MRS to confirm lymphoma.
    • rCBV ≫ 2 → GBM more likely; still do DCE for atypical cases.
  3. DCE

    • High Ktrans / extraction fraction with low vp → PCNSL.
    • Moderate Ktrans with high vp → GBM.
  4. MRS

    • Lipid/Lac peak outside necrosis + Cho/Cr > 2.6 → PCNSL.
    • Only necrotic-core lipids, Cho/Cr ~ 3, ± glycine → GBM.
  5. Integrate DWI/ADC (low ADC in PCNSL) and clinical factors (age > 60, immunosuppression).


10.5 5 | Pitfalls & Practical Tips

  • Steroids rapidly lower lymphoma rCBV & enhancement—perform imaging before high-dose steroid therapy. (PMC)
  • Hemorrhage / calcification cause susceptibility drop-outs on DSC—use spin-echo or multi-echo alternatives. (ScienceDirect)
  • Hyper-vascular PCNSL (~10 %) can mimic GBM; permeability metrics (extraction fraction, Ktrans) rescue specificity. (ajnr.org)
  • Always normalise to contralateral white matter to reduce scanner-to-scanner variability. (PMC)

10.6 6 | One-Glance Summary Table

Technique Key Metric GBM (Typical) PCNSL (Typical) Cut-off / Clue Most Common Pitfall
DSC Max/mean rCBV 3 – 6 (heterogeneous) 0.8 – 2 (uniformly low) rCBV < 1.5 ⇒ PCNSL Hyper-vascular PCNSL subset
PSR Low High High PSR (>90 %) ⇒ PCNSL Uncorrected leakage
DCE Ktrans (core) 0.2 – 0.6 0.3 – 0.8 High with low rCBV ⇒ PCNSL Low temporal resolution
Kep (rim) High Lower AUC 0.90 (Kep) ROI mis-placement
MRS Lipid/Lac peaks Necrotic only Diffuse, intense Non-necrotic lipids ⇒ PCNSL Short TE required
Cho/Cr 2 – 4 > 2.6 Threshold 2.6 Partial-volume CSF

10.6.1 Take-home

When rCBV and metabolic data disagree, believe permeability & lipids—a “leaky, lipid-rich, but low-vascular” tumour is lymphoma until proven otherwise. Conversely, a “vascular, necrotic, cho-rich” lesion points toward glioblastoma. Apply the whole toolkit systematically to maximise diagnostic confidence and to guide biopsy planning or upfront chemoradiation.